Analysis of dynamic appliance flexibility considering user behavior via non-intrusive load monitoring and deep user modeling

The research on non-intrusive load monitoring (NILM) and the growing deployment of home energy management system (HEMS) have made it possible for households to have a detailed understanding of their power usage and to make appliances participate in demand response (DR) programs. Appliance flexibility analysis helps the HEMS dispatching appliances to participate in DR programs without violating user's comfort level. In this paper, a dynamic appliance flexibility analysis approach using the smart meter data is presented. In the training phase, the smart meter data is preprocessed by NILM to obtain user's appliances usage behaviors, which is used to train the user model. During operation, the NILM is used to infer recent appliances usage behaviors, and then the user model predicts user's appliances usage behaviors in the DR period considering long-term behaviors dependences, correlations between appliances and temporal information. The flexibility of each appliance is calculated based on the appliance characteristics as well as the predicted user's appliances usage behaviors caused by the control of the appliance. The HEMS can choose the appliance with high flexibility to participate in the DR programs. The case study demonstrates the performance of the user model and illustrates how the appliance flexibility analysis is performed using a real-world case.